Open Bug 1597297 Opened 5 years ago Updated 1 year ago

The JS debugging option Async Stacks is enabled in the Talos performance profile, sometimes lowering test scores, sometimes introducing noise

Categories

(Testing :: Talos, task, P3)

Product:
Testing
Component:
Talos
Version:
Version 3
Type:
task

Tracking

(Not tracked)

People

(Reporter: mconley, Unassigned)

References

Details

(Whiteboard: [fxp])

This bug is modeled after bug 1595537, where a similar issue was investigated for the Raptor test suite.

Here's a comparison of all of our Talos tests with async stacks on (base) vs off (new): https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=af7016983c70a191c39889d9ffdce9345ebe2e55&newProject=try&newRevision=4be37d95375d3fd905f90d8296e92d93cdb438db&framework=1

There are gains in sessionrestore_many_windows opt e10s stylo, startup_about_home_paint_realworld_webextensions and tp5o_webext.

Interestingly, this also seems to drop some of the noise from our measurements on Linux and Mac OS. Unfortunately, in some cases it seems to increase the noise for Windows.

Whiteboard: [fxperf] → [fxperf:p2]
Priority: -- → P2
Severity: normal → S3
Priority: P2 → P3

Mike, can you help me understand where this fits in the perf keyword hierarchy? I'm not quite sure...

Flags: needinfo?(mconley)

I think this is a good idea to reduce Talos test noise, but ultimately isn't something our users will notice.

Flags: needinfo?(mconley)
Whiteboard: [fxperf:p2]
Whiteboard: [fxp]
You need to log in before you can comment on or make changes to this bug.